Technical Field
Embodiments of the present disclosure relate to techniques for controlling an electronic converter, such as switched mode power supplies.
Description of the Related Art
Power converters are electronic circuits that transform electrical energy from one form to another and are able to control its flow from a source to a load. Whichever type of power converter one can consider (there are ac-dc, dc-dc, ac-ac and dc-ac power converters, depending on the input and the output), the concept of control is inherent in their nature.
The block diagram of FIG. 1 shows a general structure of a power converter 20, located between a power source 10, such as a battery or the mains, and a load 30.
In the example considered, the power converter 20 includes a power stage 22, sometimes termed “power circuit”, and a control unit 24, which is connected to the power stage 22 and controls the power stage 22 operation based on measurements of a number of electrical quantities.
Accordingly, the power converter 20 receives power from the source 10 and converts the electrical energy to a different form to apply to the load 30.
The power stage 22 can be implemented by many circuit topologies, which often include a combination of power semiconductor devices and passive components, mainly transformers, inductors, and/or capacitors. In some converters, the power stage 22 may involve multiple conversion stages using the same or different topologies connected either in a cascade or in parallel.
The control unit 24 receives the measured electrical quantities through one or more sensors S1 and S2 which monitor the operation conditions of the power stage. For example, in FIG. 1 is shown a first group of sensors S1 coupled to the input lines of the power stage 22 and a second group of sensors S2 coupled to the output lines of the power stage 22. For example, the sensors S2 may monitor the output voltage Vout and/or the output current Iout, and the sensors S1 may monitor the input voltage Vin and/or input current Iin. However, in general, also other quantities, both electrical and thermal, may be taken into consideration. Based on the measured quantities received, the control unit 24 outputs control signals that enable the power stage 22 to modulate and control the energy flow, effectively regulating the desired electrical quantities.
Modern power conversion is often based on switched-mode technology, where one or more electronic switches are used to close or open branches in the power circuit 22 at a switching frequency, fSW, to control energy flow. The switches are often power semiconductor switches driven by a control unit, i.e., the control quantities output by the control unit are binary (two-level) pulsed signals that determine the open and closed state of the controllable power switches. The power switches may be any suitable semiconductor device for handling high-power switching operations, such as bipolar junction transistors (BJTs), field effect transistors (FETs), and/or insulated gate bipolar transistors (IGBTs). The switches can also include diodes.
For example, for dc-dc converters, the control unit 24 is configured to keep the dc output voltage Vout and/or the dc output current Iout constant, even under changing operating conditions. Those of skill in the art will appreciate that an ac input power may be converted in a dc input power, e.g., by means of a rectifier, such as a bridge rectifier.
For example, operating conditions may change as a result of changes in the dc input voltage Vin or changes in the power demanded by the load 30. The control unit 24 ensures that the output quantity to be regulated (Vout or Iout) is as close as possible to a preset constant value, also referred to as the setpoint.
FIG. 2 illustrates a control unit 24 having a common closed-loop, negative-feedback control design. The control unit of FIG. 2 includes four major blocks: a sensing circuit 240, an optional signal conditioning circuit 242, a modulator 244 and a driver 246.
The sensing circuit 240 measures the quantity Xout to be regulated, e.g., either the output voltage Vout or the output current Iout, and produces a measured value signal x being representative of Xout. Optionally, the sensing circuit 240 may sense other electrical quantities in the power circuit 22 that are used to perform the control action. The measured value signal x is then transmitted to the optional signal conditioning circuit 242.
The signal conditioning circuit 242 processes the measured value signal x coming from the sensing circuit 240. In particular, the signal conditioning circuit 242 receives the measured value signal x and generates a control signal y, such as a control voltage vC or a control current iC, based on the measured value signal x. Whether the control signal y is a control voltage vC or a control current iC is unrelated to whether the quantity Xout to be regulated is the output voltage Vout or the output current Iout.
For example, the signal conditioning circuit 242 may include a frequency-compensated error amplifier, such as an I (integral), PI (proportional-integral) or PID (proportional-integral-derivative) regulator. Typically such a regulator is implemented with an operational amplifier (op-amp) surrounded by a passive network that also defines its frequency response in the frequency range that is significant for the control loop (up to about fSW/2).
The modulator 244 receives the control signal y and possible other signals directly provided by the sensing circuit 240, properly conditioned if necessary. Specifically, the modulator 244 modulates a quantity Ψ, which the power stage ultimately uses to control the energy flow. In the context of switched mode power stages, the modulator 244 outputs a train of low-power two-level pulsed signals qj(t) that are received by the driver 246.
For example, the driver 246 may be a power amplifier and/or a level shifter that receives the low-power inputs qj(t) and produces the higher power signals Qj(t). The signals Qj(t) have an amplitude and a power level suitable to drive the power switches of the power stage 22.
When the operating conditions of the converter 20 change, any deviation in the regulated quantity Xout from the setpoint produces a change in x and, then, in the control signal y. This change in y results in a change in the quantity Ψ handled by the modulator 244, and this change tends to balance the input-to-output energy flow. This balance ensures that the regulated quantity Xout remains as close as possible to the setpoint.
In order to achieve an appropriate control of the output quantity Xout, the control system 24 should be designed to ensure a stable control loop, good regulation, and good dynamic performance. A stable control loop will let the regulated quantity Xout recover a steady-state value after the change in the operating conditions. Good regulation is met when the steady-state values of the regulated quantity Xout before and after the change are as close to the setpoint as possible. Finally, good dynamic performance is achieved when the regulated quantity Xout does not excessively deviate from the setpoint during the transient and the transient itself fades away in a short time.
These control objectives may be expressed in terms of characteristic quantities of the transfer function of the control loop, such as bandwidth, phase margin and dc gain. The objectives can be achieved by acting on the frequency response of the error amplifier network in the signal conditioning circuit 242, such as setting its gain and appropriately placing the poles and zeroes of its transfer function. This may be achieved by selection of the value of resistors and capacitors that make up the passive network attached to the amplifier.
The structure of the modulator 244 or, in other words, the nature of the quantity Ψ it handles, determines the method for controlling the regulation of the output quantity Xout. There are many of them. One group of methods is based on pulse-width modulation (PWM), and includes methods such as “duty cycle control,” (aka “voltage mode control”) “peak current-mode control,” and “average current-mode control”, to name the most popular ones. With the duty cycle control method the quantity Ψ is the ratio between the time TON during which a power switch is closed to the switching period TSW=1/fSW. With the peak current-mode control method the quantity Ψ is the peak current flowing through the energy storage magnetic device. With the average current-mode control method the quantity Ψ is the average current flowing through the energy storage magnetic device. With these methods the switching frequency fSW is usually fixed but not necessarily.
In addition to PWM control methods there are also pulse frequency modulation (PFM) methods, where the switching frequency fSW is variable by definition. Among the many existing methods we can mention the “direct frequency control” method, where Ψ is the switching frequency of the converter; and the “time-shift control” method, where Ψ is the amount of time from a zero-crossing of the current in the energy storage magnetic device to the next change of state of the power switches.
Another important characteristic of the power circuit that impacts on how the control circuit is implemented, in particular the way the control signal y is passed on to the modulator, is whether the converter is isolated or non-isolated. This “isolation” refers to the existence of an electrical barrier between the input and output of the converter 20.
For example, FIG. 3a shows a boost converter. Generally, a boost converter comprises two input terminals for receiving an input voltage Vin and two output terminals for providing an output voltage Vout. A boost converter is a non-isolated converter, because it has a common ground terminal GND for both the input and the output.
As well known to those skilled in the art, the positive input terminal is connected to the positive output terminal via an inductor L and an electronic switch D1, usually in the form of a diode. A further electronic switch SW1 is connected between the intermediate point between the inductor L and the diode D1, and ground GND. Finally, a capacitor Cout is usually connected in parallel with the output. The electrical connection between the input and output make non-isolated converters simple and cost efficient, but limits their usage to certain applications, such as Point-Of-Load (POL) converters.
In the example considered, a control unit 24 is used to drive the switch SW1 as a function of the output voltage Vout. Such non-isolated converters do not need any special electrical provision to provide the control signal to the modulator. If the circuits are properly combined, the output of the signal conditioning circuit can be connected directly to the modulator input.
However, especially for power converters running from the mains, many safety agency bodies or customers require a separation from the applied input voltage and the output voltage, which is often user accessible.
FIG. 3b shows in this respect that such an isolation barrier of converter may be crossed by means of a (high frequency) transformer T, which removes the direct electrical connection from the input to the output.
For example, the circuit shown in FIG. 3b is based on the flyback topology. In this case, the converter comprises at the primary side of the transformer T an electronic switch SW2, which is connected with the primary winding of the transformer T in series between the input terminals. On the secondary side, the converter comprises a flyback diode D2, which is connected with the secondary winding of the transformer T in series between the output terminals. Also in this case, an output capacitor Cout may be connected in parallel with the output.
Also in this case, a control unit 24 may be used to drive the switch SW2 as a function of the output voltage Vout. Accordingly, with such isolated converters, the power is switched on the input side (commonly referred to as the primary side), but under control from the output side (commonly referred to as the secondary side) in order to provide proper regulation. This requirement introduces an additional problem, namely that signals from the secondary side are transmitted to the primary side. The requirement for primary side switching to be controlled by secondary side characteristics requires a second connection crossing the isolation barrier.
The signal to be fed back to the primary side depends on where the modulator 244 and the driver 246 are physically located. Most commonly, both the modulator 244 and the driver 246 are located on the primary side (typically, both embedded in a control integrated circuit or IC). In this case, the control signal y is fed back to the primary side. This case is commonly referred to as “primary control”.
In other implementations all the parts of the control unit except the driver 246 are located on the secondary side (again, typically embedded in a control IC). In this case, commonly referred to as “secondary control”, the two-level pulsed signals qj(t) or Qj(t) are fed back to the primary side. No matter what signal is transferred back and although this path involves only information, rather than power, it should still be isolated.
For example, FIG. 4 shows a common solution of a signal conditioning circuit 242 configured to feed the control signal y back to the primary side in case the output quantity to be regulated is Vout (i.e., Xout=Vout).
In this arrangement a three-pin adjustable shunt regulator SR, such as a TL431, is used as secondary reference/error amplifier that drives an optocoupler OC. Basically, the shunt regulator SR is configured to sense the output voltage Vout, e.g., by means of a voltage sensor S2a comprising a voltage divider consisting in two resistors R1 and R2, and produces a control signal based on the difference between the setpoint and the actual value, while the optocoupler OC transfers the control signal to the primary side. Those of skill in the art will appreciate that the signal conditioning circuit 242 may also comprise a compensation network CN comprising, e.g., one or more capacitors and/or resistors.
Specifically, in the example considered, the photodiode of the optocoupler OC is connected with a resistor R3 and the shunt regulator SR in series between the output terminals, i.e., Vout. Accordingly, with this circuit arrangement, output voltage changes ΔVout cause corresponding changes ΔiΦ in the current iΦ flowing through the resistor R3 and the photodiode of the optocoupler OC. The current change ΔiΦ determines a proportional change Δic in the current ic drawn by the phototransistor of the optocoupler OC. This current may be used to drive the modulator 244 directly (in this case y=ic), or may be first converted into a voltage before being fed into the modulator (y=vc). For example, in FIG. 4 is shown a resistor RFB, which is connected for this purpose between the output of the optocoupler OC, i.e., the feedback pin of the modulator 244 and a constant voltage indicated with Vbus.
It is a common market requirement in many power converters to specify conversion efficiency targets over a wide range of power levels demanded by the load. To meet this goal it is necessary to take some control actions in response to this power level.
Typical examples of these actions include modifying some control parameters (e.g., the switching frequency), or changing the events determining the turn on and off of the power switch(s) (e.g., switches SW1 and SW2 shown in FIGS. 3a and 3b) or operating the converter intermittently (commonly referred to as burst-mode operation) when the power demanded by the load falls below a certain level to maximize power conversion efficiency at light load. Moreover, in multi-phase converters (i.e., with multiple power stages connected in parallel) it is also desirable to change the number of operating stages according to the power demanded by the load to optimize the conversion efficiency over a very broad power range.
In addition to these measures in favor of energy efficiency, there are also protection functions that need to be considered. A typical requirement is to limit the maximum power deliverable by the converter as a protection in case of load failures.
In converters with secondary control these tasks are relatively easy to fulfil because the control IC may have direct access to the output quantities (Vout, Iout) and elaborate them to derive the power Pout demanded by the load and take actions consequently. This case, therefore, is not interesting in this context and will not be considered any more.
In the more common case of converters with primary control, the control IC does not have direct access to the output voltage Vout and current Iout, but can only directly read the input voltage Vin and the input current Iin, thus assessing the input power to the converter Pin. The only information usually received from the secondary side is the control signal y.
There is one more issue to consider: as previously stated, to maximize power conversion efficiency at light load, converters are often required to work intermittently (burst-mode operation) and, during the idle periods when the converter is not switching, the input current Iin and, then, the input power Pin fall to essentially zero. As a result, any system computing the input power Pin through the reading of the input voltage Vin and the input current Iin can provide the information to stop the converter but cannot provide the information to restart the converter. An additional functional block would be required to provide this information.
Based on these considerations, it would be convenient to use the control signal y to perform these kinds of actions because this would lead to very simple circuit implementations thanks to a signal that may be always active regardless the power circuit is operating continuously or is temporarily stopped (e.g., during burst-mode operation). Additionally, being the control signal bounded within a range, the maximum input power would be inherently limited.
The inventors have observed that, to use the control signal y as an input power gauge, there should be a one-to-one accurate relationship between the power level Pin and the control signal y:Pin=f(y,p1, . . . pn,c1, . . . cm)  (1)where p1, . . . pn is a set of parameters characterizing the power stage and c1, . . . cm is a set of parameters relevant to the control unit. Both pi (i=1, . . . n) and cj (j=1, m) are assumed to be constant values subject to statistical spread.
Unfortunately a relationship like that does not exist for most of the known control methods. However, usually it is possible to find a relationship like:Pin=f(y,Vin,Vout,p1, . . . pn,c1, . . . cm)  (2)
This can be analyzed as follows. Generally speaking, the power level Pin is related to the quantity Φ with a relationship such as:Pin=g(y,Ψ,Vin,Vout,p1, . . . pn,c1, . . . ck)  (3)and the control signal y is related to Ψ with a relationship that can be expressed as:Ψ=h(y,Vin,ck+1, . . . cm)  (4)
The structure and the arguments of the functions g and h depend on the topology of the power stage 22 and on the control method respectively.
As to the link between Ψ and Pin, in some cases Ψ is only loosely related to the power level, in other cases it is tightly related but there may be a significant dependence on Vin and/or Vout too.
PWM-controlled dc-dc converters operated in the Continuous Conduction Mode (CCM) using the duty-cycle control method are often an example of the first case. In fact, to a first approximation, in these systems the duty cycle depends on Vin and Vout only, not on Iout (i.e., there is no Pin=g(Ψ) function). However, in the real operation, there is often a slight dependence of the duty-cycle on the power level because the switch-on time of the power switch needs to be slightly extended to compensate for power losses (which, in turn, depend on Iout).
Resonant dc-dc converters using either the direct frequency control method or the “time-shift control” method are often another example of the first case: frequency and time-shift are a weak function of the power level; they change little with the power level and are much more affected by the input-to-output voltage ratio.
PWM-controlled dc-dc converters using the average current mode control method are often an example of the second case. In fact, with this method the quantity Ψ is usually the dc input current Iin, which is strongly related to Pin; however Iin depends on the input voltage Vin too, so this method is effective to show Pin if the input voltage is fixed or varying in a narrow range.
A possible exception in this panorama may be represented by PWM-controlled buck-boost or flyback converters operated at a fixed frequency in the Discontinuous Conduction Mode (DCM) using the peak current-mode control method. In this case Pin depends mainly on the quantity Ψ (the peak current Ipk):Pin=½LIpk2fSW  (5)
The other aspect to take into consideration is the statistical spread of the parameters pi and cj. This affects both Pin=g(Ψ, . . . ) and Ψ=h(y, . . . ) and causes the quantity Ψ and, then, the control signal y to spread in a certain range for a given power level from unit to unit. Accordingly, one crucial point is the sensitivity of the functions g and h with respect to the parameters pi and cj. A couple of examples of how Pin=g(Ψ, . . . ) is affected by the tolerance of pi and cj will be now given.
For example, in a resonant converter the statistical spread of the components of its resonant tank (ϵpi) causes the switching frequency (Ψ) to be different for a given Pin and a given input-to-output voltage ratio. Additionally, the sensitivity of the switching frequency to the spread of these parameters changes considerably with the operating conditions and can go from an almost negligible to a very high level.
In the previously mentioned fixed-frequency, DCM-operated, peak-current mode controlled buck-boost or flyback converter, the statistical spread of the value L (ϵpi) of the inductor and the tolerance of the oscillator frequency, which the switching frequency fsw(ϵcj) is locked to, result in different values of the controlled peak current Ipk (i.e., Ψ) for the same power level Pin and, then, different values of the control signal y.
Again with reference to this converter, a parameter ϵcj that adversely affects the accuracy of the Ψ=h(y, . . . ) relationship is the propagation delay of the current sense comparator. Because of this delay, the controlled peak current Ipk slightly exceeds the value programmed by the control signal y; this extra current depends on the amount of this delay and on the slope of the current, which, in turn, depends on the inductance value of the converter and the input voltage Vin. As a result, the control signal y depends on Vin as well and not only on the input power Pin.
In this regards, A. S. Kislovski, “A new control principle for switching regulators”, Proceedings of PCI, September 1983, Page(s) 178-186, with a review in D. Gouttenegre, B. Velaerts, T. Michaux, “Modelling and Analysis of dc-dc Converters Control by Power Equalization”, Power Electronics Specialists Conference, 1988. PESC'88 Record, 19th Annual IEEE, Page(s) 960-967, vol. 2, proposed an “input-output power equalization” control method.
FIG. 5 shows the basic operating principle of this control method.
The dc input voltage Vin and the instantaneous input current Iin are measured. For example, in FIG. 5 is shown a current sensor S1b configured to measure the current Iin.
The input current Iin is provided to a resettable integrator 248 synchronized with a clock the signal CLK that determines also the turn-on of the power switch of the power stage 22. Specifically, the signal CLK also sets a PWM latch 250, whose output Q essentially determines the duty cycle of the power switch of the power stage 22. The output Vint of the integrator 248 will be a non-linear ramp signal that starts from zero at the beginning of each switching cycle (just after the power switch is turned on) and reaches a final value, just before being reset, that is proportional to the electrical charge taken from the power source in a switching cycle. The switching period T is assumed to be constant, thus this charge is also proportional to the average value of Iin (Īin) in a switching cycle, and such is the peak value of the ramp Vint too.
The dc input voltage Vin and Vint are provided to the inputs of a wide-bandwidth analog multiplier/divider (MD) block 252. The wide-bandwidth requirement for MD stems from the need to follow the signal Vint as closely as possible. The block 252 is provided with a third input Vx and outputs a signal i*:
                              i          *                =                                            V              in                        ·                          V              int                                            V            x                                              (        6        )            
The signal Vx is the output of a proportional-integral-derivative (PID) regulator 254 that senses the output voltage Vout and compares it against a reference voltage Vref. Vx can be regarded as the control signal y in the general schematic of FIG. 2.
The output of the MD block 252 goes to the non-inverting input of a comparator 256 that receives on its inverting input a signal proportional to the dc output current Iout. For example, in FIG. 5 is used for this purpose a current sensor S2b configured to measure the current Iout.
Under steady-state conditions Iout, Vin and Vx are constant, so the signal i* is a nonlinear ramp with the same shape as Vint but with a different amplitude, adjusted by the voltages Vin and Vx. When the ramp i* equals Iout the output of the comparator 256 goes high and resets the PWM latch 250, causing the power switch of the stage 22 to turn off.
In this way, the comparator 256 maintains the equality of the output current Iout and the peak value of i* cycle-by-cycle. Therefore:
                                                        V              in                        ·                                          I                _                            in                                            V            x                          =                  I          out                                    (        7        )            
Assuming lossless operation (Pin=Pout):Vin·Īin=Iout·Vout  (8)
it follows from equation (7) that:Vout=Vx  (9)and it is possible to state that the system in FIG. 5 maintains the output voltage Vout at the desired level by equalizing the input and the output powers of the power stage.
This power equalizing feature brings a lot of benefits in terms of dynamic behavior because it performs a fast corrective action in case of perturbations. For example, if the input voltage is perturbed, the block 252 corrects the input power within a switching period before any deviation in the output voltage Vout is observable. Similarly, if the output current Iout is perturbed (e.g., due to varying load conditions) the control readjusts the input power to suit the new power demand within a switching period before the output voltage Vout is perturbed. However, in this solution merely the driving of the power stage is adapted based on the input power, i.e., the input voltage Vin and the input current Iin. However, the feedback control signal y indicates still only the output voltage Vout and thus cannot be used as an input power gauge.